Enabling users to work sustainably on shared institute computing resources

TL;DR

The paper presents the VISPA project, a user-centric approach to enhancing sustainability in a physics computing cluster through energy monitoring, green scheduling, and behavioral incentives.

Contribution

It introduces a comprehensive system combining technological tools and behavioral strategies to promote sustainable resource use in shared research computing environments.

Findings

Per-job energy monitoring enables users to track their carbon footprint.

Green-window scheduling aligns job execution with renewable energy availability.

Voluntary measures foster a cultural shift towards sustainability.

Abstract

The VISPA project is a self-managed, mid-scale computing cluster that supports physics data analysis in research and teaching. Because the cluster is housed in a 1970s institute building with limited retrofit options, conventional efficiency upgrades would yield only minor energy savings. We therefore target sustainability primarily through user-centric measures. A monitoring system now records per-job energy consumption, while real-time data on the renewable share of the German power grid enable `green-window' scheduling. Users can query their individual energy consumption and carbon footprints, receive weekly reports, and tag jobs by project for aggregate accounting; memory records from previous runs help avoid oversubscription. All options are voluntary, fostering a cultural shift rather than imposing hard constraints. A simulation framework evaluates the potential impact of these measures. Together, the technological and behavioral interventions aim at medium- to long-term reductions in greenhouse-gas emissions by increasing resource awareness within the scientific community.